-
Notifications
You must be signed in to change notification settings - Fork 0
/
Copy pathDemo4BLS.py
149 lines (122 loc) · 5.05 KB
/
Demo4BLS.py
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
# -*- coding: utf-8 -*-
# 文件名通配符
import glob
# os操作系统的功能接口函数
import os
# numpy用于矩阵处理
import numpy as np
# PIL 图像处理包
from PIL import Image
# 下两行代码作用：遇到截断的JPEG时，程序就会跳过去，读取另一张图片
from PIL import ImageFile
ImageFile.LOAD_TRUNCATED_IMAGES = True
from BroadLearningSystem01 import BLS, BLS_AddEnhanceNodes, BLS_AddFeatureEnhanceNodes
# TRAIN_IMG_PATH表示文件夹的路径，glob.glob匹配所有的符合条件的文件，并将其以list的形式返回
TRAIN_IMG_PATH2=r'C:\Users\mlamp\Documents\03increase\crack'
TRAIN_IMG_PATH1=r'C:\Users\mlamp\Documents\03increase\damage'
Test_IMG_PATH2=r'C:\Users\mlamp\Documents\01basis\basicCrack'
Test_IMG_PATH1=r'C:\Users\mlamp\Documents\01basis\basicDamage'
all_img_paths1 = glob.glob(os.path.join(TRAIN_IMG_PATH1, "*.png"))
all_img_paths2 = glob.glob(os.path.join(TRAIN_IMG_PATH2, "*.png"))
all_img_paths01 = glob.glob(os.path.join(Test_IMG_PATH1, "*.png"))
all_img_paths02 = glob.glob(os.path.join(Test_IMG_PATH2, "*.png"))
# zeros(shape, dtype=float, order='C')
# 返回：返回来一个给定形状和类型的用0填充的数组；
# 参数：shape
# X01、X02、X03存储图像转为numpy数据时的数组 256*256=65536
X01 = np.zeros(65536)
X02 = np.zeros(65536)
X = np.zeros(65536)
# print(X01.shape)
imgs1 = []
imgs2 = []
# 将图片缩放到目标大小，不改变原数据，需将得到结果赋值给新变量
target_size = (256, 256)
# 遍历训练数据集1 将所有图片添加进一个数组
for img_path in all_img_paths1:
# 读取图片
img1 = Image.open(img_path)
# resize 重新缩放 将图片大小调整到(256,256)
new_image = img1.resize(target_size)
# 创建数组用来存放图片
X1 = np.array(new_image)
# 3通道的图片取图片数据的其中一个通道，数据是列的结构
X1 = X1[:,:,0:1]
# 将列结构转化为行结构，变成行展开的一维数组
X1 = X1.reshape(-1)
# 将每个同一标签下的图片（转为一维之后的向量）向量纵向堆叠
# 直白点就是第一次循环之后X01的结构是[[65535个0],[图片单通道的一维数组]]，循环就是以此叠加
# 所有训练集1中的数据都放入了X01
X01 = np.vstack((X01, X1))
# X01.shape返回[多少行一维度数组,每个维度的存储数字的数量]
# X01.shape[0]取出了X01中的一共有多少行一维数组（其中一行是全部0，其余行是真正的图片）
# np.ones创建全部值为1的一维数组（个数由参数决定）
# Y1的作用就是给训练数据集1 打上标签1作为记号
Y1= np.ones((X01.shape[0]))
# 可能是转置？疑问
Y1=np.transpose(Y1)
# 遍历训练数据集2 将所有图片添加进一个数组
for img_path in all_img_paths2:
img2=Image.open(img_path)
print("img2:",img2)
print("img2:",type(img2))
new_image = img2.resize(target_size)
# imgs1.append(img1)
# img_gray = color.rgb2gray(new_image)
X2 = np.array(new_image)
X2=X2[:,:,0:1]
X2=X2.reshape(-1)
print(X2.shape,X02.shape)
X02 = np.vstack((X02, X2))
# Y2的作用就是给训练数据集2 打上标签2作为记号
Y2 = np.ones((X02.shape[0]))*2
Y2 =np.transpose(Y2)
# 将2个数据集的图片一起堆积起来得到最后的X就是所有的 数据集
X = np.vstack((X01, X02))
# 将2个数据集的图片的标签一起堆积起来得到最后的Y就是所有的 标签集
Y = np.hstack((Y1, Y2))
Y = np.transpose(Y)
X012 = np.zeros(65536)
X022 = np.zeros(65536)
for img_path in all_img_paths01:
img12=Image.open(img_path)
new_image = img12.resize(target_size)
X12 = np.array(new_image)
X12=X12[:,:,0:1]
X12=X12.reshape(-1)
X012 = np.vstack((X012, X12))
Y12= np.ones((X012.shape[0]))
Y12=np.transpose(Y12)
for img_path in all_img_paths02:
img22=Image.open(img_path)
new_image = img22.resize(target_size)
X22 = np.array(new_image)
X22 = X22[:, :, 0:1]
X22 = X22.reshape(-1)
X022 = np.vstack((X022, X22))
Y22 = np.ones((X022.shape[0]))*2
Y22 = np.transpose(Y22)
# X2是所有测试集的数据
X2 = np.vstack((X012, X022))
# Y2是所有测试集的标签
Y2 = np.hstack((Y12, Y22))
Y2 = np.transpose(Y2)
traindata = X
trainlabel = Y
testdata = X2
testlabel = Y2
N1 = 10 # # of nodes belong to each window
N2 = 10 # # of windows -------Feature mapping layer
N3 = 500 # # of enhancement nodes -----Enhance layer
L = 2 # # of incremental steps
M1 = 50 # # of adding enhance nodes
s = 0.8 # shrink coefficient
C = 2**-30 # Regularization coefficient
print('-------------------BLS_BASE---------------------------')
BLS(traindata, trainlabel, testdata, testlabel, s, C, N1, N2, N3)
print('-------------------BLS_ENHANCE------------------------')
BLS_AddEnhanceNodes(traindata, trainlabel, testdata, testlabel, s, C, N1, N2, N3, L, M1)
print('-------------------BLS_FEATURE&ENHANCE----------------')
M2 = 50 # # of adding feature mapping nodes
M3 = 50 # # of adding enhance nodes
BLS_AddFeatureEnhanceNodes(traindata, trainlabel, testdata, testlabel, s, C, N1, N2, N3, L, M1, M2, M3)